A zonal approach to managing congestion has been adopted in the Norwegian scheduled power market.

The trading process works approximately as follows: 1) Based on the supply and demand schedule bids given by the market participants, the market is cleared while ignoring any grid limitations. This produces a system price of energy. 2) If the resulting flows induce capacity problems, the nodes of the grid are partitioned into zones. 3) Considering the case with two zones defined, the zone with net supply is defined as the lowprice area, whereas the net demand zone is determined the high-price area. 4) Net transmission over the zone-boundary is fixed when curtailed to meet the violated capacity limit. 5) The zonal markets are now cleared separately giving one price for each zone, being the low price and the high price. If the flow resulting from this equilibrium still violates the capacity limit, the process is repeated from step 4). If any new limits are violated the process would be repeated from step 2), possibly generating additional zones. 6) The revenue of the grid-company, (from capacity charges), is equal to the price difference times the transmission across the zone-boundary. An assumption made in the six steps given above is that a zone boundary should cut the link with the capacity problem. In a large network this still leaves the grid-company, Statnett, with a huge flexibility when defining the zone-boundaries. According to Statnett the Norwegian system can be interpreted as inflicting a positive capacity charge in the low price area and a negative charge in the high price area (relative to the system price of energy). This means that withdrawals are charged in the high price area and compensated in the low price area. For net injections the opposite is valid. As pointed out above it is not exactly clear how the number of zones and zone-boundaries are to be determined. Stoft shows that the partition of the network into zones generally is not obvious1, but states that it should be based on price differences, the reason being that the dead-weight loss resulting from erroneous prices is generally proportional to the square of the pricing error. Walton and Tabors also focus on price differentials and suggest that it might be possible to use statistical methods using the standard deviation of nodal price distributions as a criterion to determine the number of zones and which nodes belong to/do not belong to the different zones. In this paper we will show the multitude of possible cuts, representing zone-boundaries, that exists even in a small example and study the resulting welfare effects. Different zone allocations will affect both the overall efficiency and the allocation of social surplus. We will also illustrate that the partition of the network into zones based on absolute values of optimal nodal price differences does not necessarily lead to a zone system with maximal social surplus. Gaming is not considered since we assume nodal markets to be competitive when calculating the market outcomes.